topsApp

Added by Alejandra . about 1 month ago

Dear ISCE developers,

I have been running topsApp.py to process 2 Sentinel-1A images. Since I'm working at a region with poor coherence I included the properties "range looks" and "azimuth looks". I also included the "filter strength". I tested this filter with different values: 0.2, 0.3, 0.25, 0.4, 0.5 (default, as far as I know) and 0.6. You can find attached the results for 'filt_topophase.unw.geo' for the six different values of filter (shown from 0.2 to 0.6 as described). As you can see there are concentric fringes but the internal part of these fringes is very different in each test. Why do I get such different results, if I only vary the filter value? What else should I consider to obtain better results/ How could I improve these results? I also attached the 'topsApp.xml' file.

Another question is, which SRTM ISCE uses as default (when it is not provided)? SRTM 3 arcsec or SRTM 1 arcsec?

Thank you for your help!

Aleja


Replies (6)

RE: topsApp - Added by Eric Fielding about 1 month ago

Dear Aleja,

The default DEM that the ISCE topsApp.py downloads is the SRTM v3 1-arcsecond DEM.

You did not mention how you set the wrapping when you displayed your interferogram tests with MDX. The default is to use 2-pi (6.28) radians wrap. To better see phase unwrapping errors, it is better to change the color wrap to a value like 20 radians. Maybe you already did that? I can't tell from the images.

For areas of low coherence, the filter strength can make a big difference, as you found. In most cases, a filter strength of 0.4 works well. You can also change the number of range looks and azimuth looks. For noisy data, more looks usually work better, so I would suggest trying 19 range and 7 azimuth looks, instead of the 7 and 3 looks that you used.

++Eric

RE: topsApp - Added by Alejandra . about 1 month ago

Dear Eric,

many thanks for your answer!

I just tested your recommendations:

  • range looks = 19
  • azimuth looks = 7
  • filter strength = 0.4
  • color wrap = 20 radians (I didn't change this value before. This time I displayed my interferogram with this command: 'mdx.py filt_topophase.unw.geo -wrap 20'. I also tested with 7-pi (21.98) )

I attached the images to this post. One with your recommendations and a zoom of +2 and additionally (for comparison) the result with 7 and 3 looks and a filter strength of 0.4 (also with a zoom of +2).

I am now a little confused about the values of these parameters, which I should use in order to obtain 'decent' and useful results.

I would very much appreciate your comments about that.

Best regards,

Aleja

RE: topsApp - Added by Eric Fielding about 1 month ago

Dear Aleja,

You did not mention the size of the area you are plotting. I think it is quite a small area and the signal of concentric fringes is a small area with fringes that are close together. From the plots you uploaded, it seems that you will need to use a smaller number of looks because the fringes look worse with the larger number of looks (19 and 7 looks). You might try something in between such as 13 range and 5 azimuth looks. The number of looks determines how many original SLC pixels are averaged in the interferogram. Averaging more pixels reduces noise, but if the fringes are too close together then too much averaging can degrade the fringes. It seems that the coherence is low enough that it is difficult to measure this small-area signal accurately. Another thing to try is processing interferograms with shorter time intervals because that should give better coherence.

Best,
++Eric

RE: topsApp - Added by Alejandra . about 1 month ago

Dear Eric,

thank you so much for your comments. I just tested with 13 and 5 looks. The result is not much better, unfortunately.

This area is, indeed, small. I want to focus on two volcanoes in this area, which has a relatively dense vegetation. Does it make sense to expand the area a bit in order to improve the results? But it could add/incorporate more noise, doesn't it?

THank you for your help.

Best regards,
Aleja

RE: topsApp - Added by Eric Fielding about 1 month ago

Dear Aleja,

Expanding the area won't help the coherence. Your main options are to find Sentinel-1 pairs with shorter time intervals or get L-band data from JAXA ALOS-2 that will have higher coherence in vegetated areas.

++Eric

RE: topsApp - Added by Alejandra . about 1 month ago

Thanks a lot Eric for your help!

Best regards,
Aleja

(1-6/6)